Builder / Prototyper / Engineer

Steven Barnett

I prototype next-gen AI tools and harnesses — turning research ideas into working systems that push the boundary of what agents can do.

I'm a software engineer focused on agentic AI systems — the intersection of large language models, tool use, memory architectures, and autonomous reasoning. I build prototypes that test the edges of what's possible with current models.

My work sits in the gap between research papers and production systems: taking novel ideas around context engineering, agent memory, and inference-time compute and turning them into things that actually run.

Agentic Systems Context Engineering Memory Architectures Tool Use Inference-Time Compute Prototyping

Agent Harnesses

Frameworks that orchestrate LLM agents with deterministic control flow — letting AI handle intelligence while code enforces correctness.

Autonomous Scanners

Systems that continuously monitor research and surface genuinely novel insights — not just new articles, but new ideas.

Rapid Prototypes

Quick-turn experiments that validate research concepts — from BM25 deduplication pipelines to multi-stage reasoning chains.

Interested in collaborating or just want to talk about AI agent architectures?

steven.abarnett@gmail.com